ARTICLE

https://doi.org/10.1038/s42003-019-0393-7 OPEN Evidence-based guidelines for controlling pH in mammalian live-cell culture systems

Johanna Michl1, Kyung Chan Park 1 & Pawel Swietach 1 1234567890():,; A fundamental variable in culture medium is its pH, which must be controlled by an appropriately formulated buffering regime, since biological processes are exquisitely sensitive to acid–base chemistry. Although awareness of the importance of pH is fostered early in the training of researchers, there are no consensus guidelines for best practice in managing pH in cell cultures, and reporting standards relating to pH are typically inadequate. Furthermore, many laboratories adopt bespoke approaches to controlling pH, some of which inadvertently produce artefacts that increase noise, compromise reproducibility or lead to the mis- interpretation of data. Here, we use real-time measurements of medium pH and intracellular pH under live-cell culture conditions to describe the effects of various buffering regimes, − including physiological CO2/HCO3 and non-volatile buffers (e.g. HEPES). We highlight those cases that result in poor control, non-intuitive outcomes and erroneous inferences. To improve data reproducibility, we propose guidelines for controlling pH in culture systems.

1 Department of Physiology, Anatomy and Genetics, University of Oxford, OX1 3PT Oxford, UK. Correspondence and requests for materials should be addressed to P.S. (email: [email protected])

COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio 1 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7

iomedical laboratories routinely perform cell culture to Results Bproduce a cellular environment that is precisely defined, Monitoring culture medium pH under incubation. Buffers are well controlled and physiologically relevant. Among the included in culture media to control acidity, yet the ensuing main chemical variables of culture systems is the concentration pH is not routinely monitored. This becomes a quality control of H+ ions, oftentimes referred to as protons. These ions issue whenever the components of buffering are being disturbed: are present in every aqueous compartment, not least from the for example, in response to metabolic acid production, or as ionization of water. Various solutes can become protonated, a consequence of transferring media between atmospheres of + thereby establishing multiple chemical equilibria involving H different CO2 partial pressures (pCO2). The dye Phenol Red ions. Consequently, the concentration of free H+ ions is (PhR) is routinely included in the media to allow investigators not intuitive to predict, but fortuitously simple to measure to assay medium acidity14,15. Such assessment could be done (e.g. with electrodes or indicator dyes). For over a century, the ‘by eye’, but a more quantitative readout of pH is obtainable pH scale has been the reporting standard for the concentration from the PhR absorbance spectrum, which can be recorded of free H+ ions1, that is, the form that is able to protonate on plate-reader platforms with an incubator chamber (e.g. targets and post-translationally modify proteins, such as Cytation 5, BioTek). To obtain a calibration curve, freshly pre- 2–4 enzymes or receptors . The much larger pool of buffered pared standards of known pH were scanned in a CO2-free + fl H ions can, however, in uence pH through dynamic re- atmosphere to prevent the acidifying effect of CO2. Calibration 5 − equilibration . solutions had no added HCO3 , as otherwise this basic sub- There is still a widely held misconception that buffers have stance would have reacted slowly with H+ ions and then escape fi an inherent ability to set the pH of a solution to a pre-de ned as CO2 gas. Fig. 1a shows absorbance spectra of PhR in level. More accurately, in a system of one dominant buffer, bicarbonate-free Dulbecco’s modified Eagle’s medium (DMEM) pH is related to the concentration of the buffer’s protonated (D7777, Sigma-Aldrich) supplemented with 10% foetal bovine (HB) and unprotonated (B) forms, and the acid dissociation serum (FBS) and 1% penicillin–streptomycin (PS; 100 U mL−1 −1 constant (pKa): penicillin, 0.1 mg mL streptomycin), 10 mM HEPES and 10 mM MES (2-(N-morpholino)-ethanesulfonic acid), and titra- ½ŠB ted to a target pH with NaOH. To correct for pH-independent pH ¼ pK þ log : ð1Þ a ½ŠHB variables, such as light path or PhR concentration, absorbance was sampled at two wavelengths that respond differently to Consider the dissolution of HEPES buffer (4-(2-hydroxyethyl)- pH. Good resolving power is attained by rationing absorbance 1-piperazineethanesulfonic acid), typically supplied as a powder at 560 and 430 nm (Fig. 1b). The best-fit equation can then be of the free acid form. This produces an acidic solution that used to convert PhR ratio assayed in subsequent experiments must be titrated (with base, e.g. NaOH) to the desired pH; once to pH. the [B]/[HB] ratio is raised to the required level, pH will − remain stable, unless there is an additional source of acid or Setting medium pH by pCO2 and [HCO3 ]. In principle, it is base. In live-cell culture, pH disturbances are an inescapable possible to control pH with one of many commercially available consequence of metabolism and there is a general tendency for buffers, but the most physiologically relevant one is CO /HCO −. fi 2 3 media to undergo acidi cation, the extent of which is also a Incubators maintain a CO2-rich atmosphere (typically 5%) to function of medium buffering capacity. In a closed system, it can enable CO /HCO − buffering. A salt of HCO − must be included 2 3 3 + be derived mathematically (see Supplementary Note) and in the medium to balance the spontaneous H -yielding CO2 shown empirically5 that peak buffering capacity is attained hydration reaction, and stabilize pH: ’ when the buffer s protonated and unprotonated forms are ðÞ$ ÀðÞþþðÞ: ð Þ equimolar, that is, when medium pH aligns with the buffer’s CO2 gas HCO3 aq H aq 2 pKa. Many exogenous buffers are available, covering a wide pH Conveniently, pH can be titrated in the range between ~6 and = − range, including HEPES (pKa 7.3; 37 °C), PIPES (piperazine-N, ~8 by varying the concentration of HCO . Figure 1c plots the ′ = − 3 N -bis(2-ethanesulphonic acid); pKa 6.7) and MES (2-(N- relationship between pCO2, [HCO3 ] and pH measured in = 6 morpholino)-ethanesulfonic acid; pKa 6.0) . DMEM (D7777, Sigma-Aldrich) containing 10% FBS and 1% A more active means of maintaining high buffering capacity PS. To keep osmolality constant, any reduction in NaHCO3 was involves the regulation of [HB] and [B], so that their ratio matched by a compensatory rise in NaCl. Over the alkaline is kept at an optimum5. This strategy underpins the rea- range, pH reported by PhR is in very close agreement with the − son why complex organisms rely on CO2/HCO3 buffer (despite prediction of the Henderson–Hasselbalch equation (Eq. 1), = 7 − = low pKa 6.15) , and have evolved gas exchange surfaces consistent with CO2/HCO3 being the dominant buffer (pKa −1 7 (e.g. lungs) and ion transport epithelia (e.g. kidneys) to empower 6.15; CO2 solubility 0.024 M atm , i.e. 1.2 mM in 5%) . − 8 − CO2 and HCO3 homeostasis . The combination of CO2 (an However, at low [HCO ], Eq. 1 underestimates pH. This − 3 acidic gas) with HCO3 (a base) produces quantitatively the discrepancy arises because the so-called ‘intrinsic’ buffers in most important buffer in extracellular body fluids. In culture medium (such as proteins included in serum) react with H+ systems, this so-called carbonic buffer is stabilized by adding an ions generated by CO hydration, pushing the equilibrium − 2 − + amount of HCO3 salt to media and enriching the incubator's (Eq 2) towards a higher [HCO ] and lower [H ], that is, a − 3 atmosphere with CO2. Here, we relate HCO3 and CO2 with less acidic medium. This correction is derived mathematically in − pH, show how the system responds to changes in its components the Supplementary Note. Thus, the concentration of HCO3 and demonstrate how the equilibrium is affected by non-volatile required for attaining a target pH is: Âà  buffers (NVBs) added to augment buffering capacity. Further- À À : ¼ ½Š´ pHtarget 6 15 þ β ´ À : : more, we explain how certain cell culture manoeuvres may lead HCO3 CO2 10 intrinsic pHtarget 7 4 fi to poor pH control. While a number of high-pro le guidelines ð Þ relating to cell culture have been published recently9–13, they 3 do not comprehensively cover the aforementioned issues per- By best fitting the data to this equation, intrinsic β −1 taining to pH. Based on our observations, we propose guidelines buffering ( intrinsic) was estimated to be 1.1 mM pH . Alter- β for good practice in controlling pH in culture systems. natively, intrinsic can be measured empirically from the

2 COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7 ARTICLE

a b 0.7 Medium pH: 10 5.44 0.6 6.60 7.00 0.5 7.45 8.43 1 0.4

0.3 0.1 0.2 PhR absorbance

0.1 560/430 nm PhR ratio

0.0 0.01 350 400 450 500 550 600 650 456789 Wavelength (nm) Medium pH

cd8.0 9 – Eq 1 CO /HCO -free media Corrected Eq 1 2 3

7.5 8 +HCl 5% CO2 7.0 7 7.5 +NaOH 7.0 6.5 6 10% CO Medium pH Medium pH 2 6.5 Buffering capacity 6.0 6.0 5 = 1/slope 5.5 = 1.64 (±0.085) mM/pH 03691215 5.5 4 0 5 10 15 20 25 30 35 40 45 012345 [Bicarbonate] (mM) HCl or NaOH added (mM)

Fig. 1 Measuring and setting medium pH under incubation. a Absorbance spectrum of Phenol Red (PhR) in Dulbecco’s modified Eagle’s medium (DMEM) (D7777) with 10% foetal bovine serum (FBS), 1% penicillin–streptomycin (PS), 10 mM HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) plus 10 mM 2-(N-morpholino)-ethanesulfonic acid (MES), and titrated (5 M HCl or 4 M NaOH) to the indicated pH. Arrows indicate wavelengths for optimal ratiometric analysis. b pH dependence of 560/430 nm ratio, fitted to curve: pH = 8.35 + log((10.9 – ratio))/(ratio – 0.0392)). c Controlling equilibrium pH − by varying pCO2 and [HCO3 ] in DMEM (D7777) supplemented with 10% FBS and 1% PS. Dashed line plots Eq. 1. Continuous line is best fit to Eq. 3 −1 (corrected version of Eq. 1), which accounts for buffering by serum (best fit: 1.11 mM pH ). Inset replots the data at low [HCO3]. d Empirical determination of intrinsic buffering capacity of DMEM (D7777; 10% FBS/1% PS, 25 mM glucose) nominally lacking buffers; titration with either HCl or NaOH. Inverse of slope provides an estimate of buffering due to serum proteins and media salts. All measurements were repeated three times (three technical replicates each). Data are shown as mean ± SEM response of pH to the step-wise addition of acid or base acid added to a medium undergoes near-complete deprotonation, (1.6 mM pH−1; Fig. 1d). which reduces pH. Physiologically, a principal source of lactic acid is glycolytic − + 16 Stability of CO2/HCO3 buffering. In most instances, media metabolism (~1:1 lactate:H stoichiometry ), which inadver- tently reduces the pH of a finite volume of medium by reacting are prepared to a neutral or alkaline pH, and over this pH − range, the Henderson–Hasselbalch equation (1) is adequate for with its HCO3 ions. An exemplar time course of medium predicting equilibrium pH. However, the robustness of Eq. 1 acidification produced by DLD1 cells is shown in Fig. 2b for a − depends on the accuracy of pCO2 and [HCO3 ] measurements. range of starting glucose concentrations. Below a starting glucose In many instances, it may be appropriate to assume that the concentration of ~12 mM, substrate availability is rate-limiting − amount of HCO3 salt added to a medium accurately predicts for lactic acid output, measured from lactate accumulation and the final concentration of base; however, some formulations glucose consumption after 6 days of incubation (Fig. 2c). When − contain weak acids that react with HCO3 salts. Under these glucose availability was not rate-limiting (>12 mM), DLD1 cells circumstances, Eq. 1 will underestimate pH, and therefore were able to produce ~20 mM of lactic acid over a period of direct pH measurements are advocated. For example, the 6 days. The extent to which lactic acid production underpins fi addition of 22 mM of NaHCO3 to media supplemented with medium acidi cation was determined by comparing lactate lactic acid will not produce the expected pH of 7.4 due to the measurements with cumulative acid production: X titration reaction (Fig. 2a). If, instead, media contained a salt ¼À ðβ Á Δ Þ: of lactic acid (e.g. Na-lactate), then the acid-titration reaction Acid produced pH − with HCO3 will not take place, and Eq. 1 adequately approx- Here, buffering capacity (β) is the sum of intrinsic buffering imates pH (Fig. 2a). The difference in behaviour between lacti- − and CO2/HCO3 -dependent buffering (see Supplementary c acid and its conjugate base can be explained in terms of Note). Plotting the relationship between these two independent equilibria: measurements (Fig. 2d) demonstrates that in DLD1 cells, medium acidification is entirely accounted for by glycolytic $ þ þ: Lactic acid Lactate H lactic acid production. Considering the magnitude of lactic acid production, medium − Around neutral pH, this equilibrium is shifted far to the right. [HCO3 ] will invariably fall below starting levels during extended After dissolving a lactate salt, only a tiny fraction of lactate will periods of incubation. As a pre-emptive measure, many types of protonate, thus the change in pH is negligible. In contrast, lactic media are formulated to contain 44 mM NaHCO3, an excess to

COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio 3 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7

ab7.7 – 7.6 DLD1 cells in 5% CO /22 mM HCO – 5% CO2/22 mM HCO3 2 3 Lactate 7.4 7.3 7.2 Glucose (mM) added start: 7.0 25 6.9 Medium pH 12.5 6.25 Lactic acid 6.8 3.125 1.56 Medium pH at equilibrium 0 6.5 6.6 0510 15 20 234567 Lactate or lactic acid added (mM) Days in culture

c 25 25 d 25 Lactate: glucose = 1.8 :1

20 20 20 Lactate : H+ = 1:1

15 15 15

10 10 10

5 5 Lactate produced (mM) 5 Increasing glucose concentration Lactate produced (mM) Glucose consumed (mM)

0 0 0 0 5 10 15 20 25 051015 20 25 Glucose added to medium (mM) Metabolic acid produced (mM)

e 1.2 f 5% CO2 5% CO2 0% CO 1.0 2 9.0 [NaHCO3] 0.8 added 8.5 0.6 8.0 0.4 44 Caco2 7.5 33 SRB (normalized) 22 0.2 DLD1 NClH747 11 Medium pH 7.0 5.5 0.0 2.8 6.6 6.9 7.2 7.5 7.8 6.5 5 h Medium pH at start of incubation 0 6.0 − Fig. 2 The control and stability of pH in CO2/HCO3 -buffered medium. a Effect of increasing [lactic acid] or [lactate] on equilibrium pH of medium (DMEM D5030) with 22 mM NaHCO3 (10% foetal bovine serum (FBS), 1% penicillin–streptomycin (PS)) placed in 5% CO2. b Effect of metabolic lactic acid production by DLD1 cells (seeding density 4,000 cells per well, growth area 0.32 cm2 per well) on the pH of medium (DMEM D5030; 10% FBS, 1% PS) containing 0–25 mM glucose (osmotically compensated with NaCl). Error bars omitted for clarity. c Effect of varying starting glucose concentration on net glucose uptake and lactate production probed on the seventh day of incubation. 90% of glucose is metabolized to lactate. d Relationship between lactate production (measured by biochemical assay) and total acid production (calculated from the pH change and buffering capacity). Slope of 1.0 indicates that medium acidification is due to lactic acid production. e Cell growth of three colorectal cancer cell lines (seeding density 4,000 cells per well, growth area 0.32 cm2 per well) measured from protein biomass (sulforhodamine B (SRB) assay) after 6 days of incubation in DMEM (D7777; 10% FBS, 1% − PS, 25 mM glucose) over a range of starting pH attained by varying [HCO3 ] at constant 5% CO2. Data are normalized to the optimum pH derived by best fit to biphasic curve. Optimal growth is near the physiological pH of 7.4. f Effect of varying pCO2 on medium pH, mimicking the withdrawal of medium from under CO2 incubation. All experiments were repeated three times (three technical replicates each). Data are shown as mean ± SEM provide a safety margin for adequate base. However, at 5% others18, but the molecular mechanism behind this response fi CO2, such media will equilibrate at pH 7.7, which is a supra- is not well de ned. Reduced proliferation at pH >7.4 may, for physiological level that can have untoward effects on cells6. instance, relate to excessive debinding of H+ ions from sensors − ‘ ’ To demonstrate the importance of physiological [HCO3 ], cell and ionic trapping of lactate in alkaline cytoplasm. growth was studied at various levels of pH attained by varying Another factor that may contribute towards pH instability − [HCO3 ], over a period of 6 days in serum-containing medium. relates to pCO2. In vivo, most mammalian cells will be exposed Cellular growth, in the presence of 10% FBS and 1% PS, was to a tightly regulated pCO2, which helps maintain pH home- interrogated by a cytotoxicity assay based on the protein-binding ostasis. By analogy, feedback circuits in incubators are designed 17 probe sulforhodamine B (SRB) . In three colorectal cancer cell to keep pCO2 constant. This environmental constancy is, lines (NCI-H747, DLD1, Caco2), growth was optimal near pH however, not always possible with cultured cells, as various 7.4, and the effect of incubation at pH 7.7 varied, with the protocol steps may require transfers between atmospheres of strongest inhibition of growth in NCI-H747 cells (Fig. 2e). The different pCO2 (e.g. in and out of an incubator). There are two notion of an optimal pH for cell growth has been noted by implications of this. First, a medium that had been titrated in a

4 COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7 ARTICLE

CO2-free atmosphere (e.g. at the bench) will acidify upon media containing such buffers carefully considers the CO2/ 6 − placement in a CO2 incubator . Second, data collected from cells HCO3 equilibrium, which takes place under CO2 incubation. that had been withdrawn from an incubator may be influenced NVB-buffered medium titrated ‘at the bench’ to a target pH will 6 by the abrupt rise in pH . This issue could be addressed by invariably become more acidic upon placement in a 5% CO2 minimizing CO2 loss from the medium (e.g. by enclosing the incubator. For example, bicarbonate-free DMEM buffered with culture dish in a regulated atmosphere), or by superfusing with 20 mM HEPES (a widely used formulation) acidifies by over half solutions pre-equilibrated with 5% CO2. The loss of CO2 from a a pH unit upon exposure to 5% CO2 (Fig. 3a). Acidification was medium can be tracked using the time course of alkalinization less pronounced at low pH because the concentration of HCO − − 3 evoked when HCO3 -containing media equilibrated at 5% CO2 required to meet the equilibrium condition is lower. The extent is transferred into a CO -free atmosphere (Fig. 2f). The small of CO hydration could be curtailed by supplementing media 2 2 − volume of media contained in 96-well plates begin to alkalinize with HCO3 salts. This, however, produces a two-buffer system immediately, with a time constant of 2–3 h. The reverse reaction in which pH dynamics are less intuitive to predict. To demon- has a time-constant of 45 min, indicating that freshly prepared strate this instability, media were prepared with 22 mM NaHCO3 media may require an hour to equilibrate inside a CO2 incubator. and one of either HEPES, PIPES or MES. These mixtures were titrated ‘at the bench’ to near the pKa of the constituent NVB, and Effect of NVBs on the stability of medium pH. The perceived then promptly placed in a 5% CO incubator for continuous − 2 drawbacks of CO2/HCO3 , namely its volatility and weaker pH monitoring. Media prepared this way demonstrated a sub- buffering at low pH, have led to the use of exogenous NVBs such stantial degree of pH instability (Fig. 3b). The direction of pH as HEPES, PIPES and MES19. It is crucial that the preparation of drift is determined by two opposing chemical reactions (Fig. 3c):

a b 5% CO2 5% CO2 0% CO2 0% CO2 8.0 8.0 30 min 7.5 7.5 error 3’ 7.0 7.0

6.5 6.5 Medium pH Medium pH and adjust pH Add NaHCO and adjust pH 30 min 6.0 6.0 Add MES, PIPES or HEPES 5.5 5.5 add MES, PIPES or HEPES c CO2

– + + – CO2 HCO3 + H H + HCO3 CO2

Buffer HBuffer HBuffer Buffer Acidifying influence Alkalinizing influence

de4 4 Caco2 DLD1

3 3

** 2 2 ** ** SRB (cell growth) 1 ** SRB (cell growth) 1 ** Media prepared by method A Media prepared by method A Media prepared by method B Media prepared by method B – – CO2/HCO3 buffer (Fig 2E) CO2/HCO3 buffer (Fig 2E) 0 0 6.6 6.8 7.0 7.2 7.4 7.6 7.8 6.6 6.8 7.0 7.2 7.4 7.6 7.8 Starting medium pH Starting medium pH − Fig. 3 pH dynamics in media prepared with non-volatile buffers without consideration of the CO2-HCO3 equilibrium. a Medium (DMEM D7777) supplemented with non-volatile buffer (HEPES, PIPES or MES; 20 mM), 10% FBS, 1% PS, and titrated to indicated target pH (large circles). Time courses show pH dynamics (measured from PhR ratio) evoked by placing media inside 5% CO2 incubator, showing a tendency to acidify. b Medium (DMEM D7777) supplemented with 22 mM NaHCO3 plus non-volatile buffer (HEPES, PIPES or MES; 20 mM), and titrated to indicated pH before placement in 5% CO2. Time courses show pH dynamics evoked by placing media in 5% CO2 incubator, demonstrating pH instability. c Schematic of the chemical processes that underpin medium pH drifts. d pH-dependence of Caco2 cell growth (seeding density 4000 cells per well, growth area 0.32 cm2 per well) measured from protein biomass (SRB assay) after 6 days of incubation in D7777 (25 mM glucose). Media were prepared by method A (20 mM

HEPES/PIPES) or method B (20 mM HEPES/PIPES plus 22 mM NaHCO3), and placed in 5% CO2. To obtain a range of pH values, HEPES- and PIPES- buffered media were mixed in various ratios. Data are plotted as a function of assumed pH (titrated ‘at the bench’). Results compared against curve − obtained with CO2/HCO3 buffer, plotted against measured equilibrium pH (data from Fig. 2e). e Experiment performed on DLD1 cells, a more pH- sensitive line. Data are shown as mean ± SD (a, b) or mean ± SEM (d, e). All measurements were repeated three times (three technical replicates each). Statistical tests: two-sided t test (**P < 0.01)

COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio 5 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7

(i) the acidifying effect of atmospheric CO2 dissolving and Fig. 2f because of the resistive action of NVBs towards pH reacting with the medium; and (ii) the alkalinizing effect arising changes. Hence, it is possible to combine an exogenous buffer − − from the equilibration between HCO3 ions and the NVB, a slow with physiological CO2/HCO3 to improve overall buffering process that had started prior to incubation in 5% CO2. The capacity, and still attain a predictable pH (Fig. 4b). balance between these opposing processes depends on the starting To test how additional buffering affects the time course of pH, and produces an array of responses that are not intuitive to medium acidification, Caco2 or DLD1 cells were cultured in CO / − 2 predict (Fig. 3b). HCO3 -buffered DMEM supplemented with 10 mM HEPES and Poorly controlled pH, such as in the instances described 10 mM MES, prepared as described in Fig. 4a. This combination above, will impinge on the accuracy and reliability of biological of two NVBs adds a constant buffering capacity over the pH recordings. To illustrate this problem, the pH dependence of range 6 to 8, making it easier to identify biological effects in growth under 5% CO2 incubation was measured in NVB-buffered media undergoing acidification. Somewhat paradoxically, 10 mM media prepared either with or without the addition of 22 mM HEPES/MES did not meaningfully reduce medium acidification NaHCO3, according to the schemes shown in Fig. 3a, b, (Fig. 4c, d), despite the obvious increase in buffering power. respectively. Experiments were performed on Caco2 (Fig. 3d) and However, the enhanced buffering was found to increase lactate DLD1 cells (Fig. 3e), representing a weakly and strongly pH- production, implying a greater collective glycolytic rate. The sensitive line, respectively (see Fig. 2e). Growth, measured by the resulting pH time course was unaffected, because the augmented SRB assay after 6 days of culture, was plotted against the pH to buffering capacity was cancelled out by the stimulated metabolic which media were titrated ‘at the bench’ (i.e. the assumed pH). In acid production, which can be expressed mathematically as: line with previous studies20, PIPES was selected for acidic media lactic acid production and HEPES was chosen for alkaline media; to obtain a range of Change in medium pH ¼À : pH, these media were mixed in various ratios. As a control, buffering capacity growth was measured separately in media prepared with CO / − 2 HCO3 in a manner that produces a predictable starting pH These observations can be explained in terms of the inhibitory 21–23 (Fig. 1c). If the error associated with pH instability under CO2 feedback of acidity on metabolic rate . Augmented buffering incubation were negligible, then the pH dependence of growth reduces the degree of medium acidification, which is permissive measured in NVB-buffered media would be the same, irrespective for a higher metabolic rate. As expected from a simple pH- of the method of medium preparation. However, the measured operated feedback circuit, the ensuing pH time course will follow pH–growth relationship was apparently different, depending on an unchanged trajectory. When the experiment was repeated on how the medium was prepared. NVB-buffered media prepared DLD1 cells using a much higher (30 mM) concentration of without NaHCO3 (Fig. 3a; method A) yielded an apparently HEPES/MES, lactate production was still stimulated, but not to a steeper pH dependence of growth, compared to NVB-buffered degree that would offset the increase in buffering (Fig. 4e). The media prepared with NaHCO (method B). This disparity, which effects of 30 mM HEPES/MES become more evident when 3 − was more pronounced in the strongly pH-sensitive DLD1 line, starting pH is reduced (i.e. when CO2/HCO3 buffering is cannot be explained merely by the chemical presence of NVBs weaker) (Fig. 4f). Thus, it is possible to curtail medium because both methods used matching concentrations of HEPES acidification with 30 mM HEPES/MES, but less than anticipated and/or PIPES. Also, the differences do not relate to inadequate from its buffering capacity per se. Whilst these observations [HCO −] (e.g. for supplying pH-regulating proteins) because all should not be generalized to all cells, they emphasize the 3 − media inside CO2 incubators eventually accumulate HCO3 from importance of making confirmatory measurements of pH, and the spontaneous hydration of CO2. Instead, the apparent shifts in taking into consideration the biological responses to increased pH–growth relationship relate to the pH error incurred during buffering, such as metabolic stimulation. medium preparation. When placed inside a CO2 incubator, NVB- Two additional precautions must be taken when using NVBs. buffered media prepared according to method A (Fig. 3a) will The first issue relates to the titration of these buffers with acids undergo an acid-shift across the pH range. This has the effect of (e.g. HCl) or bases (e.g. NaOH), which introduces additional over-estimating the degree of growth inhibition at low pH. In osmolytes (Na+,Cl−) into the medium (Fig. 4g). The build-up of contrast, the pH of NVB-buffered media prepared with NaHCO3 osmolytes can be substantial, for example, the addition of 20 mM and titrated ‘at the bench’ (Fig. 3b) will converge towards pH ~7 HEPES and titration to pH 7.4, introduces ~30 mOsm kg−1 of during CO2 incubation. This produces apparently pH-insensitive additional osmolytes (i.e. excess of ~10%). A major increase in growth, because the test range of pH is, in reality, narrowed. osmolality would lead to cell shrinkage, changes in membrane Ultimately, the error was introduced because medium pH was set tension and potentially a myriad of downstream effects24. Indeed, − in a manner that did not take into account the CO2-HCO3 supra-physiological osmolality is likely to influence the results of equilibrium. the experiment shown in Fig. 4e, f. Some media formulations are Notwithstanding the issues described above, there may be valid available without buffers, giving some leeway for adding extra reasons to supplement media with NVBs (e.g. to limit medium osmolytes within physiological limits. For example, a total of fi −1 − acidi cation under long-term cell culture of highly glycolytic cell 88 mOsm kg of additional osmolytes can be added to HCO3 - lines)20. The apparent instability of systems containing a NVB free medium D7777 (Sigma-Aldrich) for the final solution to − plus CO2/HCO3 (described in Fig. 3) could be addressed by attain the same osmolality as ready-to-use medium D5796 modifying the protocol for preparing media. In the first step, the (Sigma-Aldrich) (Fig. 4h). The second issue to consider relates NVB should be added to bicarbonate-free media, and then to the binding properties of buffers. Although NVBs are primarily ‘ ’ − + fi 2+ titrated to a target level at the bench . To include CO2/HCO3 chelators of H ions, they also show modest af nity for Ca buffer, its components must be added at a concentration ratio ions. Lower [Ca2+] reduces the driving force for Ca2+ entry into that will be in equilibrium with the target pH (Fig. 1c). In the case cells and hence the state of Ca2+ signalling cascades25. Solutions of HEPES-buffered media titrated to pH 7.4, this would require containing 20 mM HEPES, PIPES or MES will reduce Ca2+ by – the addition of 22 mM of NaHCO3 and placement in 5% CO2. ~10 15% (Fig. 4i). Whilst this may not have a paramount effect Media prepared this way will equilibrate to the target pH within on Ca2+-dependent properties, it will contribute towards 2 h when aliquoted into 96-well plates (Fig. 4a). Note that the increased noise and weaker statistical power. This issue could equilibration takes longer than in the experiment shown in be avoided by adding CaCl2 to compensate for the chelation.

6 COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7 ARTICLE

a b 8.0 5% CO 0% CO2 2 8.4 7.6 30 min 8.0 7.2 3 7.6 6.8 7.2 Medium pH Add NaHCO

6.8 Medium pH (equilibrium) 6.4 6.4 6.8 7.2 7.6 8.0

6.4 Add HEPES/MES, adjust pH Target medium pH

c – d CO /HCO – 7.6 CO2/HCO3 35 7.6 2 3 35 +10 mM HEPES/MES +10 mM HEPES/MES 30 * 30 7.4 * 7.4 25 25 7.2 7.2 20 P = 0.45 20 P = 0.60 15 15 7.0 Caco2 7.0 DLD1 [Lactate] (mM) Medium pH Medium pH 10 10 6.8 [Lactate] (mM) 6.8 5 5 β +15% β +19% 6.6 0 6.6 0 234567End- 234567End- Days under incubation pont Days under incubation pont

e 7.6 CO /HCO – 35 f 7.4 CO /HCO – 25 2 3 ** 2 3 +30 mM HEPES/MES 30 +30 mM HEPES/MES ** 7.4 7.2 20 25 7.2 P<0.001 20 7.0 P<0.001 15 15 7.0 DLD1 6.8 DLD1 10 [Lactate] (mM) [Lactate] (mM) Medium pH 10 Medium pH 6.8 6.6 5 5 β +61% β +107% 6.6 0 6.4 0 234567End- 234567End- pont Days under incubation Days under incubation pont

gh i 20 400 2.1 15 300 ] (mM) 10 200 2+ 1.4 – 3 20 mM MES Osmolality 5 20 mM PIPES (mOsm/kg) 100 20 mM HEPES 0.7 Free [Ca D1152 D7777 mM NaOH added D5796 0 0 456789 HEPES – – 25 mM 20 mM PIPES 20 mM HEPES 22 mM HCO 20 mM MES Target pH Nacl 110 110 75 mM 0.0 NaHCO3 44 – – mM − − Fig. 4 Enhancing buffering capacity of CO2/HCO3 -containing media with non-volatile buffers, with consideration of the CO2-HCO3 equilibrium. a Medium (DMEMD 7777, 10% FBS, 1% PS, 25 mM glucose) supplemented with non-volatile buffer HEPES and MES (10 mM), and titrated to indicated target pH (large circles). NaHCO3 then added to a concentration expected to be in equilibrium with 5% CO2 at target pH (Fig. 1c). Time course of pH equilibration under 5% CO2 from different starting levels. Repeated three times (three technical replicates each). b Good agreement between target and measured equilibrium pH. c Time course of medium acidification in Caco2 cells (seeding density 4,000 cells per well, growth 2 − − area of 0.32 cm per well). Media buffered with 5% CO2/22 mM HCO3 , or the combination of CO2/HCO3 plus 10 mM HEPES/MES (a 19% increase in time-averaged buffering, β). Medium lactate accumulation at the end point was greater with enhanced buffering. d Experiment performed on DLD1 cells. e Experiment performed on DLD1 cells with 30 mM HEPES/MES. f Experiment repeated from a more acidic starting pH, at which 30 mM HEPES/MES is expected to provide half of total buffering. Measurements repeated four times (three technical replicates each). Statistical tests: lactate measurements tested by two-sided t test (*P < 0.05, **P < 0.01); time courses tested by two-way analysis of variance (ANOVA) (P value for the effect of buffering is stated). g Increasing buffering capacity with non-volatile buffers also increases osmolarity due to the buffer molecules and the base required for titration. Calculated [NaOH] required to titrate MES, PIPES or HEPES buffer to a target pH. h Osmolality of three different media − formulations. Arrows show gap in osmolality in HCO3 -free media, which can be filled with buffer and acid–base required for titration, plus additional NaCl required to bring osmolality to a physiological level. Note, in the case of medium D7777 and D1152, a total of 88 and 132 mOsm kg−1 can be added, respectively. i Free [Ca2+] measured by electrode, showing partial Ca2+ chelation by non-volatile buffers. Data are shown as mean ± SEM. Repeated three times

COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio 7 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7

High-throughput analysis of intracellular pH. A fundamental pHi in a monolayer. These pHi data can be obtained by loading reason why changes in medium pH (controlled or unwarranted) cells with pH-sensitive fluorescence dye cSNARF127. To identify fl can in uence cellular physiology is because intracellular pH (pHi) the centroids of cells, nuclei can be stained with Hoechst-33342, responds to changes in extracellular pH (pHe). This coupling which is spectrally resolvable from cSNARF1. After a period arises because the proteins that regulate pHi are also sensitive to of loading (10 min) and wash-out in dye-free media, stacks of – fl fl pHe, and a re-balancing of transmembrane acid base uxes images were collected, corresponding to 447 nm uorescence alters steady-state pHi. A major contributor to these acid–base excited at 377 nm (Hoechst-33342) and of 590 and 640 nm fl − fl fl uxes are HCO3 transporters, which are active only in the uorescence excited at 531 nm (cSNARF1) (Fig. 5a). Of ine, the − 8,26 presence of CO2/HCO3 buffering . Thus, an assessment of the pHi in individual cells was inferred from the cSNARF1 fluores- effects of medium pH and buffering regime on cell behaviours cence probed around nuclei, identified by particle analysis of should consider actions mediated through changes in pHi. Plate- Hoechst-33342 images. After ratioing background-subtracted fl fl based imaging platforms allow high-throughput uorescence uorescence at 590 and 640 nm, pHi can be sampled individu- measurements that can capture the population distribution of ally for each cell (Fig. 5a). An example of a suitable code, written

a cSNARF-1 Hoechst INTRACELLULAR pH b 1.6

1.2

0.8

590 nm/64 nm ratio 0.4

0.0 5 6789 1 mm 6.9 7.4 Intracellular pH

c d – Caco2 DLD1 CO2/HCO3 – CO2/HCO3 HEPES/MES 0.3 0.3 (CO /HCO –free) pHe=7.4 HEPES/MES pHe=7.4 2 3 – (CO2/HCO3 free) Frequency Frequency 0.0 0.0 6.8 7.0 7.2 7.4 7.6 6.8 7.0 7.2 7.4 7.6 Intracellular pH Intracellular pH

ef7.6 Caco2 7.6 DLD1

– CO /HCO – 7.4 CO2/HCO3 7.4 2 3 HEPES/MES HEPES/MES – (CO /HCO –free) (CO2/HCO3 free) 2 3 7.2 7.2

7.0 7.0 Intracellular pH Intracellular Intracellular pH Intracellular 6.8 6.8

6.6 6.6 6.0 6.3 6.6 6.9 7.2 7.5 7.8 6.0 6.3 6.6 6.9 7.2 7.5 7.8 Extracellular pH Extracellular pH

Fig. 5 Effects of buffering regime and medium pH on intracellular pH, measured using a high-throughput imaging method. a Monolayer of DLD1 cells imaged with Cytation 5 plate reader. Image on left shows superimposition of cSNARF1 and Hoechst-33342 fluorescence maps. Image on right shows pH in individual cells, identified by nuclear staining. b Calibration curve determined from nine colorectal cancer cell lines (LS174T, PMFKO14, LS513, HCT15, SW620, GP2D, HCT116, Caco2, RW2892; seeding density 100,000 cells per well, growth area 0.56 cm2 per well). Three technical replicates each. Best-fit curve: pH = 6.978 + log((1.497 −ratio)/(ratio − 0.221)). c Histogram of intracellular pH in Caco2 monolayers bathed in D7777-based media (25 mM − glucose) at pH 7.4, buffered by either 5% CO2/22 mM HCO3 , or 10 mM HEPES + MES titrated to 7.4 (CO2 free). Note the substantial alkalinization in the absence of physiological buffer. Repeated three times (three technical repeats each). d Experiment performed on DLD1 monolayers. e Effect of medium pH − − on intracellular pH in Caco2 monolayers. The pH of CO2/HCO3 -buffered media was varied by changing [HCO3 ] (incubation in 5% CO2). In contrast, − the pH of HEPES/MES-buffered media were titrated to target pH with NaOH at the bench (incubation in 0% CO2). Best fit: linear (CO2/HCO3 )or − polynomial (HEPES/MES). Note that intracellular pH is more responsive to changes in extracellular pH in the absence of physiological (CO2/HCO3 ) buffer. f Experiment repeated on DLD1 monolayers. Data are shown as mean ± SD. Repeated three times (three technical repeats each)

8 COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7 ARTICLE as a MATLAB script, is included as Supplementary Code 1. This In a retrospective review of articles published in Nature and approach was first applied to generate a calibration curve with Cancer Research (third quartile of 2017, a period selected at the nigericin method28, in which cells are incubated in solutions random for the purpose of this analysis) reveals that only a small containing 100 µM nigericin (a H+/K+ ionophore), 140 mM KCl percentage of studies provide the necessary information about the + (to balance intracellular K ), 0.5 mM EGTA, 1 mM MgCl2, buffering regime and pH of culture media. Three-quarters of 10 mM MES (for pH <7) or 10 mM HEPES (for pH >7) titrated articles published in Cancer Research and two-thirds of life to a desired target pH in a CO2-free atmosphere. The science articles published in Nature present data from cultured calibration curve shown in Fig. 5b was determined in CO2-free cells. However, just under half of these articles report the man- conditions, and the best-fit equation can be applied to convert ufacturer of the medium, and only a tenth give information about measured fluorescence ratio into pH for cells under various test the buffer composition. Only a third of all studies report the conditions. pCO2 in incubators: typically 5%, although some using 10% CO2 This pH imaging approach was used to investigate the effect (which then necessitates a proportional adjustment to HCO −). A i − 3 − of CO2/HCO3 buffering on pHi. Figure 5c, d show histograms of significant number of studies use media containing [HCO3 ] pH in populations of Caco2 or DLD1 cells, incubated either in outside the range 22–26 mM, producing a non-physiological i − media buffered with 5% CO2/22 mM HCO3 (equilibrated at pH pH. Among the studies that reported the use of 5% CO2, 7.4), or CO /HCO −-free media buffered by 10 mM HEPES and approximately one-third used classical DMEM, the underlying 2 3 − − 10 mM MES (titrated to pH 7.4). In the absence of HCO3 ions, formulation of which contains 44 mM HCO3 (which would the pHi of DLD1 and Caco2 cells was shifted in the alkaline equilibrate to pH 7.7 in 5% CO2). Less than a tenth of studies direction by 0.3 units, which is attributable to the inactivation of used media supplemented with HEPES, half of which were a − fi pHi-regulating HCO3 transporters. The direction and magni- mixture of HEPES and bicarbonate, which are identi ed herein as tude of this effect is likely to be cell type-dependent, and therefore potentially problematic (Fig. 3). not intuitive to predict. By repeating these experiments over a Attaining a fine degree of control over pH is realistically – range of media pH, it is possible to map the pHe pHi relationship achievable in modern culture systems, and efforts should be made (Fig. 5e, f). The pHi of cells was more sensitive to changes in pHe to implement the best practice in a bid to improve the accuracy, − fl in the absence of CO2/HCO3 . For example, pHi in Caco2 cells compatibility and reproducibility of measurements. The ow fi − acidi ed by twice as much in the absence of CO2/HCO3 in chart in Fig. 6b illustrates the suggested steps in setting the pH response to a drop in pHe from 7.4 to 6.4 (Fig. 5e). Since the of media. Supplementary Data 1 provides further details of these majority of H+ targets are intracellular, such buffer regime- steps, illustrated in a selection of media from a major supplier. – dependent changes in pHe pHi coupling can lead to erroneous Based on the observations described herein, we make the fol- inferences concerning the mechanisms of cell responses to lowing recommendations: − microenvironmental acid–base challenges. Recommendation 1: CO2/HCO3 is the physiological buffer and therefore should be the preferred choice for biological research. Its use avoids possible unwarranted effects that Discussion exogenous buffers may have19, for example, longer-term Research in virtually every biomedical laboratory relies on cell toxicity33–35,Ca2+ binding (Fig. 4i) or glycolytic stimulation – culture, either to maintain cells in a state that is conducive (Fig. 4c f). Including CO2 as part of the buffering regime also for physiologically relevant activity or to explore the effects of establishes a more realistic transmembrane [CO2] gradient for controlled chemical, physical or biological influences. Cultured those cells that generate CO . Additionally, the presence of − 2 cells will remain an essential biological resource, offering a HCO3 ions activates essential membrane transport processes tractable model for characterizing pathways, recording responses responsible for cellular pH homeostasis8,26. This influences and manipulating disease-related processes29. However, the steady-state intracellular pH and its sensitivity to changes in translational relevance of findings borne from culture systems is extracellular pHe (Fig. 5c–f), an important transduction critically dependent on the extent to which in vitro conditions mechanism by which medium pH modulates cellular behaviours. relate to in vivo setting. Furthermore, the value of any experi- Recommendation 2: media exposed to an atmosphere enri- mental finding is determined by its reproducibility. However, ched in CO2 must include an appropriate concentration − 70% of scientists surveyed recently by Nature were unable of HCO3 salt in order to stabilize at the required pH. CO2/ ’ 30 − to reproduce another s experiment , and a post-publication HCO3 is unusual among buffers because its acidic component analysis has suggested a reproducibility rate of as little as 10% in is a gas. Consequently, precautionary measures are warranted 31 − cancer biology . The inadequate quality of preclinical data has when handling CO2/HCO3 -buffered media in open chambers been linked to the high failure rate of agents progressing from to avoid the loss of gas, and hence alkalinization. Whilst this in vitro validation to phase III testing, of the order of 95% in chemical peculiarity is desirable in vivo because it allows the cancer research32. One factor contributing towards these out- lungs to regulate buffering, it poses a challenge for experiments comes has been attributed to variables relating to environmental involving changes in ambient pCO2. Media that are to be exposed conditions31. to a raised pCO (e.g. inside a CO incubator) must contain a − 2 2 Commercial sources of media offer a wide range of formula- salt of HCO3 in order for the buffer to promptly stabilize at tions, summarized in Fig. 6a for three major types: DMEM, a predictable pH. Conveniently, medium pH could be set by MEM and RPMI-1640. Fewer than half of the available for- changing the ratio of pCO to [HCO −] (Fig. 1c). When taking − 2 3 mulations contain physiological [HCO3 ], and a substantial readings outside CO2 incubators, it is important to consider number of options include media with considerably lower or the pH dynamics associated with abrupt shifts in pCO . The − 2 higher [HCO3 ], which would produce acidic and alkaline con- pH of media in small volumes (e.g. 200 µL) will begin to rise ditions, respectively (Fig. 2a). Special precautions are needed immediately when removed from a CO2 incubator, and may with various formulations supplemented with HEPES because require hours to attain the new equilibrium. Conversely, when these may produce unexpected pH dynamics inside CO incu- preparing media for incubation, adequate time should be allowed − 2 bators (Fig. 3). Formulations that lack CO2/HCO3 and any for equilibration inside a CO2 incubator. This mitigates the risk major NVB provide are useful starting point for producing of an unwarranted transient alkaline stimulus imposed on cells bespoke media (Fig. 4). by an out-of-equilibrium medium.

COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio 9 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7

Raised a = 1 medium Physiological bicarbonate bicarbonate (~30 mm) (22–26 mM) DMEM MEM RPMI

High No buffer bicarbonate added (<44 mM)

HEPES

Low bicarbonate (<20 mM)

b Identify major buffers... NVBs, bicarbonate? Estimate osmolality... ~2×[NaCl]+2×[NaHCO3]+2×[KC]+[Glucose]

Start – β Calculate target [HCO3 ]... Eq 1, or Eq 3 if intrinsic is known

Are both Are NVBs Are NVBs Is medium Reduce – – NVBs and No the only No required in No [HCO3 ] Yes [HCO3 ] by – HCO3 major medium? greater than titraton with present? buffer? Yes target HCl – [HCO3 ] Medium No Likely to should not Is total Top-up with produce contain osmolality NaCl to unstable pH – Supplement Yes HCO3 medium above physio- under CO2 No incubation with salt of physio- logical HCO – logical? osmolality Add NVB 3 Yes Hyperosmotic Titrate to target pH with NaOH or HCl at 37 °C Add serum (if medium may cause required) and unwarranted place in CO2 effects on cells incubator

Fig. 6 Summary of buffering regimes in commercially-available media formulations, and flow chart showing instructions for preparing media at a target pH. a Venn diagrams summarizing the commercial availability of Dulbecco’s modified Eagle’s medium (DMEM), minimum essential medium (MEM) or RPMI- 1640 buffers (supplied by Sigma-Aldrich and Thermo Fisher Scientific), grouped by buffering regime. Area is proportional to the number of media available − in each category. Media with physiological HCO3 and no additional non-volatile buffer are highlighted with a thick black border. HEPES-buffered media are indicated with a red outline. b Flow chart guiding through the steps required to adjust the pH of culture media. NVB: non-volatile buffer (e.g. HEPES). Target − [HCO3 ] for a given medium pH can be calculated from Eq. 3. Total osmolality can be approximated as 2 × [NaCl] + 2 × [NaHCO3] + 2 × [KCl] + [Glucose]. See Supplementary Data 1 for further details of these steps

Recommendation 3: media supplemented with non-volatile Recommendation 4: reporting standards must provide ade- buffers may have unstable pH if these are prepared without quate information about the buffering regime. This should − consideration of the CO2-HCO3 equilibrium. For periods include a description of buffer composition, CO2 partial pressure outside CO incubators, non-volatile buffers can be used to and, in the case of bespoke media, the steps involved in 2 − stabilize pH, provided that HCO3 salts are not included preparing media. (to match the absence of CO2). Conversely, for experiments involving CO incubation, non-volatile buffers should not be 2 − used in lieu of HCO3 , as this results in media becoming Methods more acidic than anticipated. If there is a good biological Cell lines and culture conditions. Human colorectal adenocarcinoma cells − Caco2, DLD1 and NCI-H747 cells were obtained from Prof. Walter Bodmer’s reason to supplement physiological CO2/HCO3 with a non- collection at the Weatherall Institute of Molecular Medicine (University of volatile buffers, the medium should first be prepared with the Oxford, UK). Caco2 and DLD1 cells were cultivated in DMEM (Life technologies, non-volatile buffers, titrated to the target pH, and then supple- Cat. No. 41965-039) supplemented with 10% FBS (Sigma-Aldrich) and 1% PS − (100 U mL−1 penicillin, 100 µg mL−1 streptomycin; Sigma-Aldrich). NCI-H747 mented with a combination of CO2 and HCO3 that is expected fi to be at equilibrium with the target pH. Some ready-made media, cells were cultivated in RPMI-1640 (Thermo Fisher Scienti c, Cat. No. 21875-034), in 5% CO2 and at 37 °C. Alternatively, cells were treated with media based on containing mixtures of several buffers, may not be compatible NaHCO3-free DMEM (Sigma-Aldrich, Cat. No. D7777), supplemented with with this sequence, and thus yield unstable pH dynamics under various concentrations of NaHCO3, NaCl, HEPES, PIPES or MES, as indicated in figure legends or NaHCO and glucose-free DMEM (Sigma-Aldrich, Cat. No. CO2 incubation. Additionally, when preparing bespoke media 3 2+ D5030) supplemented with various concentrations of glucose and NaCl, as with non-volatile buffers, changes in free [Ca ] and total indicated in figure legends. Lines were authenticated by single nucleotide poly- osmolality must be considered to avoid non-physiological morphism (SNP)-based profiling and tested routinely for mycoplasma conditions. contamination.

10 COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7 ARTICLE

Monitoring medium pH using absorbance. Medium pH was measured by PhR 7. Leem, C. H. & Vaughan-Jones, R. D. Out-of-equilibrium pH transients absorbance at 430 and 560 nm using Cytation 5 imaging plate reader (Biotek) in the guinea-pig ventricular myocyte. J. Physiol. 509(Pt 2), 471–485 equipped with a CO2 gas controller (Biotek). Measurements were taken from (1998). 200 µL medium in clear, flat-bottom 96-well plates (Costar) with lids at 37 °C. 8. Boron, W. F. Regulation of intracellular pH. Adv. Physiol. Educ. 28, 160–179 Media were based on NaHCO3-free DMEM (Sigma-Aldrich, Cat. No. D7777), (2004). supplemented with 10% FBS, 1% PS and various concentrations of NaHCO3, 9. Price, P. J. Best practices for media selection for mammalian cells. In Vitro Cell NaCl, HEPES, PIPES or MES, as indicated in figure legends. Alternatively, media Dev. Biol. Anim. 53, 673–681 (2017). based on NaHCO3-free, glucose-free DMEM (Sigma-Aldrich Cat. No. D5030) 10. Coecke, S. et al. Guidance on good cell culture practice. a report of the second supplemented with 10% FBS, 1% PS and various concentrations of glucose and ECVAM task force on good cell culture practice. Altern Lab. Anim. 33, fi NaCl, as indicated in gure legends was used. 261–287 (2005). 11. Pamies, D. et al. Good cell culture practice for stem cells and stem-cell-derived models. ALTEX 34,95–132 (2017). Cell growth analysis using SRB. Cells were plated in triplicates at densities of 12. Pamies, D. et al. Advanced good cell culture practice for human primary, stem 4,000 cells per well in clear, flat-bottom, 96-well plates with a growth area of 0.32 cell-derived and organoid models as well as microphysiological systems. cm2 per well (Costar). The following day, the medium was replaced with 200 µL 35 – DMEM (Sigma-Aldrich, Cat. No. D7777), supplemented with 10% FBS, 1% PS and ALTEX , 353 378 (2018). 13. Geraghty, R. J. et al. Guidelines for the use of cell lines in biomedical research. various concentrations of NaHCO3, NaCl, HEPES, PIPES or MES, as indicated in fi Br. J. Cancer 111, 1021–1046 (2014). gure legends. Alternatively, media based on NaHCO3-free, glucose-free DMEM (Sigma-Aldrich Cat. No. D5030) supplemented with 10% FBS, 1% PS and various 14. Clark, W. M. & Lubs, H. A. The colorimetric determination of hydrogen ion – concentrations of glucose and NaCl, as indicated in figure legends was used. Cells concentration and its applications in bacteriology: III. J. Bacteriol. 2, 191 236 (1917). were cultured for 6 days and pHe was monitored on each day using PhR absor- bance. After 6 days, the cells were fixed using 10% trichloroacetic acid at 4 °C for 15. Haas, A. R. Colorimetric determination of the hydrogen ion concentration in small quantities of solution. J. Biol.Chem. 38, 49 (1919). 60 min. Afterwards, they were washed with H2O four times, and stained with SRB (0.057% in 1% acetic acid) for 30 min. Residual SRB was removed by 16. Vinnakota, K. C. & Kushmerick, M. J. Point: muscle lactate and H(+) washing four times with 1% acetic acid. SRB was then dissolved in 10 mM Tris production do have a 1:1 association in skeletal muscle. J. Appl. Physiol. (1985) base. SRB absorbance was read at 520 nm absorbance using Cytation 5 imaging 110, 1487–1489 (2011). discussion 1497. plate reader (Biotek). 17. Vichai, V. & Kirtikara, K. Sulforhodamine B colorimetric assay for cytotoxicity screening. Nat. Protoc. 1, 1112–1116 (2006). 18. Eagle, H. The effect of environmental pH on the growth of normal and pHi measurements. Cells were plated in triplicate at 100,000 cells per well in malignant cells. J. Cell. Physiol. 82,1–8 (1973). black wall, flat coverslip bottom µ-plate 96-well plates with a growth area of 19. Ferguson, W. J. et al. Hydrogen ion buffers for biological research. Anal. 0.56 cm2 per well (Ibidi) and were left to attach overnight. They were then – − Biochem. 104, 300 310 (1980). incubated in media supplemented with cSNARF1-AM (5 µg mL 1, Molecular 20. Damaghi, M. et al. Chronic acidosis in the tumour microenvironment selects −1 Probes) and the nuclear stain Hoechst-33342 (10 µg mL , Molecular Probes), for overexpression of LAMP2 in the plasma membrane. Nat. Commun. 6, fl for 10 min, and then replaced with dye-free medium (twice). Images of uores- 8752 (2015). fl cence excited at 377 nm and collected at 447 nm (Hoechst-33342), and of uor- 21. Wilmes, A. et al. Towards optimisation of induced pluripotent cell culture: escence excited at 531 nm and collected at 590 nm and 640 nm (cSNARF1), extracellular acidification results in growth arrest of iPSC prior to nutrient were acquired using Cytation 5 imaging plate reader and its bespoke software. – − exhaustion. Toxicol. In Vitro 45, 445 454 (2017). For media buffered with CO2/HCO3 , measurements were performed in an 22. Erecinska, M., Deas, J. & Silver, I. A. The effect of pH on glycolysis and atmosphere of 5% CO2, established in the plate reader. Further analysis of the phosphofructokinase activity in cultured cells and synaptosomes. population distribution of pH data was performed with a MATLAB script J. Neurochem. 65, 2765–2772 (1995). (Supplementary Code 1). 23. Hu, X., Chao, M. & Wu, H. Central role of lactate and proton in cancer cell resistance to glucose deprivation and its clinical translation. Signal. Transduct. Lactate and free calcium measurements. Free lactate and calcium Target Ther. 2, 16047 (2017). concentrations were determined using ABX Pentra 400 (Horiba) from 150 µL 24. Pedersen, S. F., Hoffmann, E. K. & Novak, I. Cell volume regulation in aliquots of medium. epithelial physiology and cancer. Front. Physiol. 4, 233 (2013). 25. Clapham, D. E. Calcium signaling. Cell 131, 1047–1058 (2007). 26. Thomas, R. C. Cell growth factors. Bicarbonate and pHi response. Nature 337, Reporting summary. Further information on experimental design is available in 601 (1989). the Nature Research Reporting Summary linked to this article. 27. Buckler, K. J. & Vaughan-Jones, R. D. Application of a new pH-sensitive fluoroprobe (carboxy-SNARF-1) for intracellular pH measurement in small, isolated cells. Pflugers Arch. 417, 234–239 (1990). Data availability 28. Thomas, J. A., Buchsbaum, R. N., Zimniak, A. & Racker, E. The datasets generating and analysed in this study are available for download as Intracellular pH measurements in Ehrlich ascites tumor cells utilizing Supplementary Data 2. spectroscopic probes generated in situ. Biochemistry 18, 2210–2218 (1979). 29. Wilding, J. L. & Bodmer, W. F. Cancer cell lines for drug discovery and development. Cancer Res. 74, 2377–2384 (2014). Received: 20 November 2018 Accepted: 14 March 2019 30. Baker, M. 1,500 scientists lift the lid on reproducibility. Nature 533, 452–454 (2016). 31. Begley, C. G. & Ellis, L. M. Drug development: raise standards for preclinical cancer research. Nature 483, 531–533 (2012). 32. Hutchinson, L. & Kirk, R. High drug attrition rates—where are we going wrong? Nat. Rev. Clin. Oncol. 8, 189–190 (2011). References 33. Hanrahan, J. W. & Tabcharani, J. A. Inhibition of an outwardly rectifying 1. Sorensen, S. P. L. Enzymstudien. II. Mitteilung. Über die Messung und die anion channel by HEPES and related buffers. J. Membr. Biol. 116,65–77 Bedeutung der Wasserstoffionenkoncentration bei enzymatischen Prozessen. (1990). Biochem. Z. 21, 131–394 (1909). 34. Lepe-Zuniga, J. L., Zigler, J. S. Jr. & Gery, I. Toxicity of light-exposed HEPES 2. Srivastava, J., Barber, D. L. & Jacobson, M. P. Intracellular pH sensors: media. J. Immunol. Methods 103, 145 (1987). fi – − design principles and functional signi cance. Physiology (Bethesda) 22,30 39 35. Stea, A. & Nurse, C. A. Contrasting effects of HEPES vs HCO3( )-buffered (2007). media on whole-cell currents in cultured chemoreceptors of the rat carotid 3. Schonichen, A., Webb, B. A., Jacobson, M. P. & Barber, D. L. Considering body. Neurosci. Lett. 132, 239–242 (1991). protonation as a posttranslational modification regulating protein structure and function. Annu. Rev. Biophys. 42, 289–314 (2013). 4. White, K. A. et al. Cancer-associated arginine-to-histidine mutations confer a Acknowledgements gain in pH sensing to mutant proteins. Sci. Signal. 10, https://doi.org/10.1126/ We thank Professor Sir Walter F. Bodmer for supplying cell lines, advice on cell scisignal.aam9931 (2017). culturing methods and for providing comments on the manuscript. We also thank 5. Roos, A. & Boron, W. F. Intracellular pH. Physiol. Rev. 61, 296–434 Dr. Alzbeta Hulikova for providing comments on how to improve the manuscript. (1981). P.S. wishes to thank Richard Vaughan-Jones for mentorship and instilling an 6. Eagle, H. Buffer combinations for mammalian cell culture. Science 174, enthusiasm for pH. The work was supported by the European Research Council, 500–503 (1971). SURVIVE #723997.

COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio 11 ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-019-0393-7

Author contributions Open Access This article is licensed under a Creative Commons

J.M. performed the experiments, K.C.P. performed the pHi experiment and undertook Attribution 4.0 International License, which permits use, sharing, the literature review, P.S. designed the research and wrote the paper. adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Additional information Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Supplementary information accompanies this paper at https://doi.org/10.1038/s42003- indicated otherwise in a credit line to the material. If material is not included in the 019-0393-7. article’s Creative Commons license and your intended use is not permitted by statutory Competing interests: The authors declare no competing interests. regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Reprints and permission information is available online at http://npg.nature.com/ licenses/by/4.0/. reprintsandpermissions/ © The Author(s) 2019 Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

12 COMMUNICATIONS BIOLOGY | (2019) 2:144 | https://doi.org/10.1038/s42003-019-0393-7 | www.nature.com/commsbio